380
1 Introduction
Variability is a fundamental property of our climate system. Two decades
ago Mitchell (1976) proposed a theoretical framework of climate variability
based on a schematic spectrum of climate variations spanning time scales
from one hour to some billions of years. He distinguishes two different types
of processes in the climate system: (i) internal stochastic mechanisms; and
(ii), external forcing mechanisms including their resonant amplification of
internal modes. The spectrum in Mitchell's original figure (Fig. 1) thus
consists of a background continuum on which is superimposed a series of
spectral peaks. The power of the background continuum is stronger for
lower frequencies and is the result of a number of red-noise spectra with
increasing characteristic time scales. According to Mitchell this is the
consequence of the stochastic aspect of the different climate system components for which first-order autoregressive processes are good conceptual
models.
The purpose of this paper, which is an extended and updated version of
Stocker (1995), is to give an overview on recent results that indicate the
importance of these processes on time scales of many decades to centuries.
Apart from a general interest this time scale is of particular importance,
because detection of anthropogenic climate change depends on our knowledge of the time scales and patterns associated with the natural level of
variability on the decadal to century time scale. A second purpose is to
summarize those mechanisms of interdecadal-to-century variability that
are quantitatively capable of causing detectable fluctuations. Up to now,
our knowledge about these processes comes mostly from a hierarchy of dynamical models of the climate system and still only marginally from an
observational or proxy network. The development and expansion of the
latter is an important task of future research.
Many aspects of Mitchell's schematic picture can be found in records of
climatic data both, directly observed and proxy. For instance, the red noise
background of the spectrum (i.e. longer time scales exhibit stronger spectral
power) can be seen in observed data from the atmosphere and ocean, and
model simulations exhibit very similar spectral characteristics (Delworth
et al. (1993), their Fig. 2, reproduced here as Fig. 5a). The spectra of
longer time series (oxygen isotopes on planktonic foraminifera shells found
in sea sediments) also confirm the red-noise continuum background and the
presence of preferred time scales of variability (see e.g. Imbrie et al. (1992,
1 Introduction
Variability is a fundamental property of our climate system. Two decades
ago Mitchell (1976) proposed a theoretical framework of climate variability
based on a schematic spectrum of climate variations spanning time scales
from one hour to some billions of years. He distinguishes two different types
of processes in the climate system: (i) internal stochastic mechanisms; and
(ii), external forcing mechanisms including their resonant amplification of
internal modes. The spectrum in Mitchell's original figure (Fig. 1) thus
consists of a background continuum on which is superimposed a series of
spectral peaks. The power of the background continuum is stronger for
lower frequencies and is the result of a number of red-noise spectra with
increasing characteristic time scales. According to Mitchell this is the
consequence of the stochastic aspect of the different climate system components for which first-order autoregressive processes are good conceptual
models.
The purpose of this paper, which is an extended and updated version of
Stocker (1995), is to give an overview on recent results that indicate the
importance of these processes on time scales of many decades to centuries.
Apart from a general interest this time scale is of particular importance,
because detection of anthropogenic climate change depends on our knowledge of the time scales and patterns associated with the natural level of
variability on the decadal to century time scale. A second purpose is to
summarize those mechanisms of interdecadal-to-century variability that
are quantitatively capable of causing detectable fluctuations. Up to now,
our knowledge about these processes comes mostly from a hierarchy of dynamical models of the climate system and still only marginally from an
observational or proxy network. The development and expansion of the
latter is an important task of future research.
Many aspects of Mitchell's schematic picture can be found in records of
climatic data both, directly observed and proxy. For instance, the red noise
background of the spectrum (i.e. longer time scales exhibit stronger spectral
power) can be seen in observed data from the atmosphere and ocean, and
model simulations exhibit very similar spectral characteristics (Delworth
et al. (1993), their Fig. 2, reproduced here as Fig. 5a). The spectra of
longer time series (oxygen isotopes on planktonic foraminifera shells found
in sea sediments) also confirm the red-noise continuum background and the
presence of preferred time scales of variability (see e.g. Imbrie et al. (1992,